The Design And Implementation Of Crowdsourced Testing Report Fusion System Based On Knowledge Graph

With the rapid development of the Internet,software technology has been applied to all aspects of life,and testing technologies are also emerging in endlessly.Crowdsourced testing refers to a test method carried out on the Internet,where the employer hires crowdsourced workers to test the specified target and submit the test results within a specified time period.Due to the large number of participants and lack of communication among workers,crowdsourced testing have the characteristics of high repetition of reports and redundant descriptions.The reports submitted by workers in the crowdsourced testing cannot be directly delivered,and specific personnel are required to identify similar reports,organize the content,and generate the final delivery report.In order to accurately identify similar reports and improve the efficiency of report compilation,this thesis designs a crowdsourced testing report fusion system based on knowledge graph.Based on the crowdsourced testing platform,this thesis designs and implements a crowdsourced testing report fusion system based on the knowledge graph.The system innovatively constructs a crowdsourced testing knowledge graph,and uses knowledge graph technology to assist report fusion.This thesis uses Natural Language Processing technology to extract the entities and relationships in the report description,and introduces the classification knowledge graph to supplement,explores the semantic connection between the reports,uses translation model to perform vector translation of the knowledge graph,and aggregates the reports describing similar defects in the same cluster.At the same time,this thesis calculates the image features in the report to assist in report fusion.Secondly,for the problem of description redundancy,this thesis uses the Page Rank algorithm to calculate the weight of the report in the report cluster,extracts the main report in the report cluster,and disassembles the report in the cluster into a single supplementary item of text or picture,and extracts the content not described in the main report forms supplementary points.This thesis also identifies the report description,checks whether there is any content that contradicts the description,and generates ambiguity points for the contradictory content.This thesis first aggregates similar reports within the task into report clusters,and then splits the report clusters into main report,description supplementary points,and description ambiguity points.This makes the classification and information acquisition of defect reports more efficient,thereby improving the efficiency of reviewer.This thesis is mainly divided into: knowledge graph module,knowledge graph report fusion module,picture calculation module and report summary module.In order to achieve efficient access to the system,the system uses Nginx for load balancing,Redis as a cache,and Thrift framework to provide efficient communication between cross-language modules.In response to the call of Xinchuang's domestic software,this thesis also carries out portability adaptation for domestic operating systems.At present,the system has been applied to related online projects.The system has realized the identification of similar defect reports in crowdsourced testing,classified and extracted the opinions in the reports,and improved the efficiency of reviewer to identify similar reports and information acquisition.The quality of delivered reports has also been effectively improved.